Abstract
Laryngeal videoendoscopy is one of the main tools in clinical examinations for voice disorders and voice research. Using high-speed videoendoscopy, it is possible to fully capture the vocal fold oscillations, however, processing the recordings typically involves a time-consuming segmentation of the glottal area by trained experts. Even though automatic methods have been proposed and the task is particularly suited for deep learning methods, there are no public datasets and benchmarks available to compare methods and to allow training of generalizing deep learning models. In an international collaboration of researchers from seven institutions from the EU and USA, we have created BAGLS, a large, multihospital dataset of 59,250 high-speed videoendoscopy frames with individually annotated segmentation masks. The frames are based on 640 recordings of healthy and disordered subjects that were recorded with varying technical equipment by numerous clinicians. The BAGLS dataset will allow an objective comparison of glottis segmentation methods and will enable interested researchers to train their own models and compare their methods.
Highlights
Background & SummaryDisorders of the human voice have a devastating impact on the affected and society in general
With the Benchmark for Automatic Glottis Segmentation (BAGLS), we aim to fill this gap and, in a collaboration of seven research groups from the USA and Europe, we created a benchmark dataset of high-speed videoendoscopy (HSV) recordings for glottis segmentation
We provide an interface to the individual data to preview and select a subsection of the data at https://www.bagls.org
Summary
Disorders of the human voice have a devastating impact on the affected and society in general. With the Benchmark for Automatic Glottis Segmentation (BAGLS), we aim to fill this gap and, in a collaboration of seven research groups from the USA and Europe, we created a benchmark dataset of HSV recordings for glottis segmentation This multihospital dataset comprises recordings from a diverse set of patients, disorders and imaging modalities. We provide the BAGLS dataset to other research groups openly online and hope that it will fuel further advances and support international collaboration in the voice, medical imaging and machine learning community Overall, this dataset fills the gap caused by the overall lack of a publicly available dataset for glottis segmentation and can serve as a litmus test for future methods for the task
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have